This isn't a better-benchmarks problem because the problem isn't that they're optimizing specific instructions that only the benchmarks use, instead they're loo

This isn't a better-benchmarks problem because the problem isn't that they're optimizing specific instructions that only the benchmarks use, instead they're looking at the name of the application and cranking up the speed if the name matches a known benchmarking application to make it look like non-benchmark apps can run faster than they really can. This is fraud IMO, regardless of if the GPU industry has been doing it for years. Being well-adjusted to a profoundly sick society is no measure of health.